Accommodative accuracy by retinoscopy versus autorefraction spherical equivalent or horizontal meridian power

Abstract

Background:

To assess agreement between accommodative lag by monocular estimation method (MEM) retinoscopy and Nott retinoscopy compared to open-field autorefraction using spherical equivalent versus power in the 180 meridian for both children and adults.

Methods:

Twenty-six children aged 7–16 years (mean: 9.9 ± 2.3) and 27 adults aged 22–29 years (mean: 24.2 ± 1.7) participated. Accommodative lag was measured by examiners with autorefraction and separate examiners using MEM and Nott retinoscopy while subjects viewed 6/18 letters at 33 cm. Five measures of autorefraction were averaged with vector analysis and both power in the 180 meridian and spherical equivalent was determined. Two-factor repeated measures analysis of variance and the mean difference and 95 per cent limits of agreement were calculated.

Results:

Mean (standard deviation) lag for each technique was: MEM = 0.69 (0.52) D, Nott = 0.62 (0.51) D, autorefraction in 180 = 0.66 (0.50) D and autorefraction spherical equivalent = 0.60 (0.46) D. Lag did not vary across techniques (p = 0.48), but children did have smaller lags than adults (p < 0.001) and greater amounts of uncorrected astigmatism (0.61 ± 0.09 D versus 0.42 ± 0.08 D, p = 0.02). There was no significant interaction between age group and technique (p = 0.74). Mean differences between techniques were small, ranging from −0.14 to +0.06 D. Ninety-five per cent limits of agreement ranged from ±0.80 to ±1.33 around the mean with the narrowest ranges found for comparisons made to autorefraction in 180. Limits of agreement were also narrowest in children as compared to adults with similar mean differences between age groups.

Conclusions:

This study demonstrates the mean agreement between autorefraction and retinoscopic techniques is centred on zero (no bias) in both children and adults. The range of agreement becomes narrower when autorefraction power in the 180 is calculated, even for a sample of subjects with moderately small amounts of uncorrected astigmatism.

Accommodation is a change in shape of the crystalline lens to focus for near viewing; however, the accommodative response is rarely precisely matched to the near demand. In most cases, the accommodative response is less than the demand of the stimulus,¹ which is termed a lag of accommodation. Accommodative lag between 0 and +0.75 D is considered normal in children and young adults for a typical near working distance (approximately 33–40 cm).^2,3 Accommodative lag typically increases beyond this range with advancing age and increasing accommodative demand.^3–5

Assessment of accommodative accuracy can be helpful in the evaluation of accommodative and binocular vision disorders. While accommodative accuracy is typically assessed with the refractive error corrected, there are instances when it may be performed unaided, such as the case of a child with uncorrected hyperopia for whom the need for refractive correction is being considered. A larger than normal accommodative lag may be associated with conditions such as accommodative insufficiency, convergence excess⁶ or poor compensation for refractive error related to uncorrected hyperopia. Less commonly, the accommodative response may be greater than the near stimulus demand, termed a lead of accommodation, and may be associated with conditions such as accommodative excess or convergence insufficiency.⁶

There is a variety of clinical methods available to measure accommodative accuracy, with monocular estimation method (MEM) retinoscopy and dynamic Nott retinoscopy being two common techniques adopted in clinical practice.⁷ While both of these methods are considered objective because the examiner determines the accommodative accuracy based on the retinoscopic reflex, there is still a potential for bias and inaccuracies in the measurement given that the examiner interprets the retinoscopic reflex to determine the accommodative lag. The open-field autorefractor offers an objective alternative to the retinoscopy techniques by eliminating any subjective judgment by the examiner, although it is not readily available in all clinical or clinical research settings.

Several studies have been conducted to assess the agreement between these tests with a summary of reported mean differences (measurement bias), standard deviations (measurement variability) and 95 per cent limits of agreement (LOA) (±1.96 * standard deviation), when available, summarised here. McClelland and Saunders found no significant difference between Nott and open-field autorefraction (mean difference = 0.06 ± 0.51 D; 95 per cent LOA not reported due to a linear increase in the differences with increase in measurement magnitude),⁸ nor did Rosenfield et al., when comparing autorefraction to Nott (mean difference = −0.02 ± 0.33 D; 95 per cent LOA = ±0.65 D) and MEM (mean difference = 0 ± 0.47 D; 95 per cent LOA = ±0.91 D).⁹

Goss et al. found larger mean differences of 0.51 D between both MEM and Nott versus spherical equivalent values obtained from Canon autorefraction (95 per cent LOA = ±0.82 D and ±0.59 D, respectively), but reported minimal mean differences when comparing Nott retinoscopy to only the spherical values from Canon autorefraction (mean difference = 0.04 ± 0.32 D; 95 per cent LOA not reported).¹⁰

In general, these studies reported good agreement between retinoscopic and autorefraction techniques, but one limitation is that all of these studies were conducted on adult subjects with the exception of one (McClelland and Saunders) that included subjects from six to 43 years (n = 38), but did not evaluate agreement with MEM.⁸

One recent study investigating agreement of retinoscopic versus autorefraction measures of accommodative accuracy in children was performed by the Pediatric Eye Disease Investigator Group (PEDIG) in 2009.¹¹ The study was designed specifically to compare the ability of MEM and Nott retinoscopy (using standard clinical methodology) to identify children whose accommodative lag measured ≥ 1.00 D by the gold standard autorefraction technique (using the same methodology employed in a previous clinical trial). The study reported mean differences of 0.40 ± 0.76 D when comparing MEM to autorefraction and differences of 0.52 ± 0.70 D when comparing Nott to autorefraction. The 95 per cent LOAs were not reported due to an observed linear relationship between the magnitude of the differences and the magnitude of the measurement. The conclusion of the paper was that neither MEM nor Nott retinoscopy had adequate sensitivity or specificity to identify children with ≥ 1.00 D accommodative lag,¹¹ thus casting doubt on the agreement of retinoscopy and autorefraction techniques in children.

However, several methodological limitations described by the authors of the PEDIG study may have impacted the outcome. First, MEM and Nott retinoscopy were performed while the patient viewed the target binocularly, while autorefraction measures were taken monocularly with one of the eyes of the subject occluded. Second, retinoscopy measurements were taken under moderate room illumination using a printed Welch Allyn card as the target, while autorefraction measures were taken with room lights off with a self-illuminated set of 6/30 letters as the target. Lastly, instead of performing all techniques through the habitual correction of the subject, autorefraction measures were taken through a trial lens that corrected the spherical equivalent refractive error of the subject, rather than matching the sphere and cylinder correction used for the retinoscopic techniques.

Given the limitations of the PEDIG study, a re-evaluation of these techniques in children that standardises the stimulus and viewing conditions across techniques is warranted. Past studies have found good agreement in adult subjects and there is no inherent reason to believe agreement would not be similarly good in children when conditions known to affect the accommodative response are equated across techniques.

In addition, it is hypothesised that previously reported differences in the agreement between retinoscopic techniques and autorefraction could be related to the common practice of measuring accuracy in only one meridian for the retinoscopic techniques (for example, the horizontal meridian) versus measurement of both sphere and cylinder power with autorefraction (often expressed as spherical equivalent for calculation of accommodative accuracy). In subjects with residual or uncorrected astigmatism, comparing a single meridian versus a spherical equivalent across techniques would inherently result in differences in measurement of accommodative accuracy. This is a possible explanation for the poorer agreement seen in the study by Goss et al.,¹⁰ as well as the PEDIG study, which included subjects with up to 1.50 D of astigmatism, but only used spherical equivalent refractions for autorefraction.¹¹

This study aims to evaluate the agreement between retinoscopic and autorefraction techniques of accommodative accuracy in both children and adults under standardised viewing conditions. Also, a comparison will be made in respect of agreement for spherical equivalent autorefraction versus calculated power in the 180 meridian to measures from the horizontal meridian with MEM and dynamic Nott retinoscopy.

Methods

This study was approved by the University of Houston Committee for the Protection of Human Subjects and followed the tenets of the Declaration of Helsinki. Written informed consent was obtained from all adult participants and parental permission and subject assent obtained for all participants under the age of 18 years.

Subjects

Subjects were recruited from the University of Houston College of Optometry and University Eye Institute. Inclusion criteria included age between seven to 40 years and ability to cooperate for study measurements. Exclusion criteria consisted of significant corneal scarring or cataracts, history of cataract surgery, near acuity worse than 6/15, presence of strabismus at distance or near, or uncorrected astigmatism greater than 1.25 DC. All subjects, or parents in the case of subjects less than 18 years, completed a brief medical questionnaire regarding any past ocular surgical history and medication use. Medication use was collected not in an attempt to exclude subjects, but rather to identify subjects taking medications that could potentially impact accommodation for further interpretation of the data. In recruiting subjects, attempts were made to enrol similar numbers of children (seven to < 18 years of age) and adults (18–40 years of age).

Preliminary tests were completed on all subjects and included distance and near acuity, distance and near cover test, and distance auto-refraction with the Grand Seiko WR-5100K (RyuSyo Industrial Co., Ltd., Kagawa, Japan) which was performed both aided and unaided for subjects presenting with refractive corrections. Lensometry was performed on spectacles if they were worn by subjects during testing and contact lens powers were documented via subject report.

Study measurements

Accommodative accuracy was measured using three techniques: Nott retinosocopy, MEM retinoscopy and autorefraction. A total of four examiners were utilised, with two examiners alternately performing MEM and Nott retinoscopy for scheduled subjects and two separate examiners alternately performing preliminary tests and autorefraction measures. The order of the three tests of accommodative accuracy was randomised and the MEM and Nott retinoscopy examiners were masked to the outcome of the autorefraction measures.

For all accommodation measures, subjects viewed a Welch Allyn Grade 6 MEM card containing nine rows of words (equivalent foot size at 40 cm = 6/18) placed at a 33 cm viewing distance (3.00 D accommodative demand) in full room illumination. All measurements were taken through the habitual correction of the subjects. Subjects were asked to read the letters in the words individually (that is, ‘spell the words’) out loud while measurements were taken. This strategy minimised head movements, particularly for autorefraction, since subjects could say each letter with minimal jaw movement.

All retinoscopic measurements were taken through a hole in the centre of the card which was aligned with the measured eye. Measurements were obtained on the right eye of each subject under both binocular and monocular viewing conditions (order randomised). As expected, greater overall lags were observed under monocular viewing conditions; however, only binocular viewing results are reported here given that the overall trends in agreement were similar between monocular and binocular viewing and measurement of accuracy is most commonly performed under binocular viewing in the clinical setting.

Nott retinoscopy

For the Nott retinoscopy technique, the stimulus was suspended at 33 cm in front of the right eye of the subject from a near rod attached to a phoropter adjacent to the subject (that is, the subject was not viewing through the phoropter, but the phoropter was positioned such that the near rod was parallel to the frontal plane of the subject to hold the target in front of the subject). While the subject viewed the card binocularly and spelled out the words on the card, the examiner observed the retinoscopic reflex in the horizontal meridian of the right eye through a hole in the centre of the card. The examiner started as close to the target as possible and moved away from the subject if ‘with’ motion was observed. Once a neutral response was found, the examiner extended a string tied around the retinoscope head to measure the distance from the point of neutrality to the stimulus card in centimetres. In cases where ‘against’ motion was observed (indicating a lead of accommodation), the examiner moved just to the side of the card to identify the point in front of the card at which neutrality was observed.

In order to calculate the final dioptric accommodative lag value, the distance from the card in centimtres was added to 33 cm if the location of neutrality was on the examiner side of the stimulus, or subtracted from 33 cm if the location of neutrality was on the subject side of the stimulus. The inverse of the distance in metres represented the total accommodative response. The response in dioptres was subtracted from the target demand of 3.00 D and expressed as the accommodative lag or lead (positive values = lag; negative values = lead).

Monocular estimation method retinoscopy

For the MEM retinoscopy technique, the stimulus was secured to the retinoscope by a magnetic strip and the 33 cm viewing distance held constant by having the subject hold a knot tied in a string attached to the retinoscope next to their right eye. The examiner would position themselves such to keep the string taut throughout the measurement. The examiner then observed the retinoscopic reflex in the horizontal meridian of the right eye of the subject as the subject spelled out the words on the card while viewing the card with both eyes open. A trial-lens bar consisting of 10 lenses ranging from −0.50 D to +2.00 D in 0.25 D steps was used to neutralise the retinoscopic reflex, with additional loose lenses used in the few instances where this range was not sufficient (full range: −1.50 to +2.50).

The reflex was initially assessed without lenses. If ‘with’ motion was observed, the +0.25 D lens was briefly introduced in front of the eye following the traditional method.⁷ Additional lenses (incremented in +0.25 D steps) were briefly introduced in front of the right eye until the first neutral response was observed. If an ‘against’ motion was initially seen, the −0.25 D lens was introduced in the same manner. The lens power was increased in −0.25 D steps until a neutral response was found. Lenses were not introduced past the initial neutral response. The lens power that resulted in a neutral response was equivalent to the accommodative lag or lead (positive values = lag; negative values = lead).

Autorefraction

Five repeated measures of refraction were taken on the right eye of each subject as they read letters on the Welch Allyn Grade 6 MEM card suspended on a near rod at 33 cm with both eyes open. The instrument settings used during the study were: static mode, vertex distance of 0 mm, and sphere/ cylinder in 0.12 D steps with one degree axis steps.

To average refraction values, the five individual refractions were first converted to M, J₀ and J₄₅ vectors,¹² averaged, and reconverted to sphere, cylinder and axis format. This average refraction was then expressed in two different ways for comparison to the retinoscopic techniques: spherical equivalent (½ cylinder power added to the spherical power) and power in the 180 meridian using the following equation, where S represents the sphere component of the subject’s average autorefraction measurement, C represents the average cylinder component, and $\phi (\text{phi)}$ (phi) is the average distance of the axis from the 180 meridian in degrees:¹³

$${\text{P}}_{\text{m}}=\text{S}+\text{C*}si{n}^{\text{2}}(\phi )$$

To express both the spherical equivalent refraction and the refraction for the 180 meridian as accommodative lag, signed refraction values were added to the target demand of 3.00 D. Resultant positive values represent an accommodative lag and negative values an accommodative lead. For example, a patient with an average autorefraction measure of −2.25 − 0.50 × 180 would have the following accommodative lags by each calculation:

Spherical equivalent:

−2.50 D + 3.00 D = + 0.50 D lag

Power in the 180 meridian:

−2.25 D + 3.00 D= +0.75Dlag

Data analysis

A two-factor repeated measures analysis of variance (ANOVA) was used to compare the magnitude of accommodative lag measured by each technique and across age groups (children versus adults). Difference versus mean analysis was performed to assess agreement between measurement techniques.¹⁴ Differences between individual measures were plotted versus the average of the measures for each pair of techniques compared and linear regression of the data performed to determine whether the magnitude of differences between techniques was related to the average magnitude of the accommodative lag.

In the absence of a significant linear relationship between differences and mean lag, the overall mean difference represents the bias between techniques and the 95 per cent LOA (calculated as 1.96 standard deviations around the mean) represents the overall agreement.¹⁴

Lastly, a descriptive analysis was performed to quantify the percentage of subjects with differences between tests falling within ±0.50 D to enable comparison to the PEDIG study which previously reported agreement using a 0.50 D criterion.¹¹

Results

A total of 57 subjects were recruited for the study. Two subjects were excluded for astigmatism greater than −1.25 DC through their habitual correction and one subject was excluded due to an incorrect autorefractor setting used during testing. Although not a pre-defined exclusion criterion, one additional subject was excluded from analysis due to reported wear of overnight orthokeratology which could result in refractive variability.

Data from 53 subjects were used in the final analysis and are presented here. There were 26 males and 27 females with a mean age of 17.1 ± 7.5 years. For analysis, subjects were divided as children (n = 26) aged seven to 16 years old (mean = 9.9 ± 2.3 years) and adults (n = 27) aged 22–29 years old (mean = 24.2 ± 1.7 years). Age binning was performed in order to address the question of whether agreement between tests was acceptable in both children and adults, rather than treating age as a continuous variable to investigate whether agreement changes as a function of increasing age. The absence of subjects from 17 to 21 years of age in the sample was not due to intentional exclusion of these individuals, but rather a reflection of the limited number of individuals in this age range employed at or visiting the college.

Subjects who presented with refractive corrections wore them for accommodative testing. The breakdown of refractive corrections worn for study measures was as follows: eight children and eight adults wore spectacles, 10 adults and one child wore soft contact lenses, and nine adults and 17 children wore no correction.

Refractive error classification

Although refractive error classification was not used for analysis, uncorrected refractive error distributions, as determined from unaided distance autorefraction, are shown in Table 1 to provide an overview of the subjects included in the study. Myopia was defined as < −0.50 D in the most plus meridian, with simple myopia having < 0.75 DC and compound myopia having ≥ 0.75 DC. Hyperopia was defined as ≥ +1.00 D in the most plus meridian, with simple hyperopia having < 0.75 DC and compound hyperopia having ≥ 0.75 DC.

Table 1.

Refractive error classifications of subjects

Refractive classification	Total no. of adults (no. wearing correction)	Total no. of children (no. wearing correction)
Emmetropia	7(1)	8(0)
Simple hyperopia	1 (0)	1 (0)
Simple myopia	11 (11)	3(1)
Astigmatic	2(0)	7(3)
Compound hyperopic astigmatism	0(0)	2(1)
Compound myopic astigmatism	6(6)	3(2)
Mixed astigmatism	0 (0)	2 (2)

Subjects were classified as astigmatic if they had ≥ 0.75 DC, but neither principal meridian fell under the myopic or hyperopic classification, whereas subjects with mixed astigmatism had one meridian that fell under the classification of hyperopia and the other under myopia.

Emmetropia was defined as both meridians ≥ −0.50 D and < +1.00 D with < 0.75 DC. The mean uncorrected spherical equivalent refractive error of all subjects combined was +0.08 ± 0.53 D (range = −1.18 to+1.94 D).

Group comparison of accommodative accuracy techniques

Measurements between techniques were compared by two-factor repeated measures ANOVA. On average for all subjects combined, accommodative lag did not vary across techniques (p = 0.48); however, children did have significantly smaller accommodative lags than adults (p < 0.001), as seen in Table 2. There was no significant interaction between age group (adults versus children) and technique (p = 0.74).

Table 2.

Group mean + standard deviation (SD) accommodative lag by technique and age group

Technique	Mean ± SD (range) Accommodative lag (D) Full group (n = 53)	Mean ± SD (range) Accommodative lag (D) Adults (n = 27)	Mean ± SD (range) Accommodative lag (D) Children (n = 26)
MEM	0.69 ± 0.52	0.89 ± 0.37	0.49 ± 0.58
	(−0.75 to+1.50)	(0 to+1.50)	(−0.75 to+1.25)
Nott	0.62 ± 0.51	0.85 ± 0.35	0.38 ± 0.55
	(−1.35 to+1.33)	(0 to+1.33)	(−1.35 to+1.17)
Autorefraction	0.60 ± 0.46	0.75 ± 0.52	0.44 ± 0.32
spherical equivalent	(−0.30 to+1.70)	(−0.30 to +1.70)	(−0.26 to+1.29)
Autorefraction	0.66 ± 0.50	0.86 ± 0.48	0.45 ± 0.43
power in the	(−0.63 to+1.91)	(−0.12 to +1.91)	(−0.63 to+1.44)
180 meridian
Overall	0.64 ± 0.50	0.84 ± 0.43†	0.44 ± 0.48†
	(−1.35 to+1.91)	(−0.30 to +1.91)	(−1.35 to+1.44)

^†Two-factor analysis of variance, p < 0.001 for age group comparison. MEM: monocular estimation method.

Differences between accommodative accuracy techniques

Mean differences between retinoscopic and autorefraction techniques for all subjects combined were close to zero (range = −0.10 to +0.04 D), as shown in Figure 1 and Table 3, and there was no significant linear relationship between the average magnitude of accommodative lag and the difference between measures for any of the full group comparisons (p ≥ 0.36).

Figure 1.

Difference versus mean plots comparing measurement techniques for all subjects combined (solid symbols children, open symbols adults) for the four comparisons (A–D) between retinoscopic and autorefraction techniques. LOA: limits of agreement, MEM: monocular estimation method, SE: spherical equivalent.

Table 3.

Summary of mean differences reported in comparing retinoscopic and autorefraction measures of accommodative accuracy

Study	n	Age	Test distance	Tests compared	Mean difference	LOA centred around the mean difference
Present study	53	7–29	33 cm	MEM versus AR SE	−0.10 ± 0.60	−1.27 to 1.07
				MEM versus AR 180	−0.03 ± 0.51	−1.02 to 0.96
				Nott versus AR SE	−0.03 ± 0.61	−1.23 to 1.17
				Nott versus AR 180	0.04 ± 0.52	−0.99 to 1.06
Present study	26	7–16	33 cm	MEM versus AR SE	−0.06 ± 0.52	−1.08 to 0.96
				MEM versus AR 180	−0.04 ± 0.42	−0.86 to 0.78
				Nott versus AR SE	0.05 ± 0.53	−0.99 to 1.09
				Nott versus AR 180	0.06 ± 0.41	−0.74 to 0.86
Present study	27	22–29	33 cm	MEM versus AR SE	−0.14 ± 0.67	−1.45 to 1.17
				MEM versus AR 180	−0.03 ± 0.59	−1.78 to 1.72
				Nott versus AR SE	−0.10 ± 0.68	−1.43 to 1.23
				Nott versus AR 180	0.01 ± 0.62	−1.21 to 1.23
Goss et al.	50	20–35	40 cm	MEM versus AR SE	0.51 ± 0.42	−0.31 to 1.33
				Nott versus AR SE	0.51 ± 0.30	−0.08 to 1.10
				Nott versus AR Sphere	0.04 ± 0.32	Not reported
McClelland and Saunders	38	6–43	25 cm	Nott versus AR	0.06 ± 0.51	Not reported
Rosenfield et al.	24	22–30	40 cm	MEM versus AR	0 ± 0.47	−0.91 to 0.91
				Nott versus AR	−0.02 ± 0.33	−0.67 to 0.63
Manny et al.	168	8–11	33 cm	MEM versus AR	0.40 ± 0.76	Not reported
				Nott versus AR	0.52 ± 0.70	Not reported

AR: autorefraction, LOA: limits of agreement, MEM: monocular estimation method, SE: spherical equivalent.

The 95 per cent LOA for all subjects combined ranged from ±0.80 to ±1.33 D for comparisons between retinoscopic and autorefraction techniques (Table 3). Mean differences and 95 per cent LOA were also calculated for children and adults separately, as shown in Table 3. In general, mean differences were similarly close to zero between age groups, although 95 per cent LOA were narrower in the children.

In comparing autorefraction spherical equivalent and power in the 180 to retinoscopic techniques, 95 per cent LOA were narrower when comparing measures in the same meridian rather than spherical equivalent (Table 3). A greater percentage of children had differences of 0.50 D or less when power in the 180 was compared (MEM: 81 per cent versus 54 per cent, Nott: 77 per cent versus 65 per cent). Improvement was also observed when using power in the 180 compared to Nott retinoscopy in adults, but not when comparing autorefraction to MEM (MEM: 67 per cent versus 70 per cent, Nott: 67 per cent versus 52 per cent).

Given the larger percentage of children falling within ±0.50 D agreement when calculating power in the 180 versus using spherical equivalent autorefraction, the distribution of uncorrected astigmatism present during accommodative accuracy testing was compared between children and adults. As seen in Table 4, a larger number of children had uncorrected astigmatism > 0.25 DC as compared to adults based on distance autorefraction through presenting corrections (unaided if no correction was worn).

Table 4.

Astigmatism magnitude measured by autorefraction over each subject’s presenting correction (habitual correction or unaided)

Uncorrected astigmatism	No. of children	No. of adults
0.00 (DC)	0	2
0.12–0.25 (DC)	2	11
0.37–0.50 (DC)	12	5
0.62–0.75 (DC)	7	6
0.87–1.00 (DC)	2	3
1.12–1.25 (DC)	3	0
Overall mean i SD (DC)†	0.61 i 0.09	0.42 i 0.08

^†t-test, p = 0.02.

The mean uncorrected astigmatism during accommodative accuracy testing in children (0.61 ± 0.09 D) was statistically significantly higher than adults (0.42 ± 0.08 D) (t-test, p = 0.02). Of the 38 subjects with > 0.25 DC, 14 were classified as with the rule (axis 001–030 and 150–180), 18 were classified as against the rule (axis 060–120) and six were classified as oblique orientation (axis 031–059 and 121–149) using previously defined criteria for astigmatism orientation.¹⁵

The presence of uncorrected oblique astigmatism did not appear to systematically affect test agreement, as only one of the six subjects with oblique astigmatism had test agreement worse than 0.50 D across all comparisons.

Potential impact of subject medications

Six subjects reported taking medications known to affect accommodation. The accommodative lags of these subjects are shown in Table 5. While the limited sample precludes formal analysis, the data are summarised here for complete disclosure given that these subjects were not excluded from analysis.

Table 5.

Accommodative lag for subjects taking medication that could impact accommodation

Subject	Lag by MEM (D)	Lag by Nott (D)	Lag by autorefraction power in the 180 (D)	Lag by autorefraction spherical equivalent (D)
Subject 1†	1.00	1.28	0.59	0.50
Subject 2†	0.50	0.65	0.32	−0.26
Subject 3†	−0.25	−0.13	0.12	0.38
Subject 4‡	−0.25	−0.33	0.34	0.20
Subject 5‡	0	0	0.50	0.56
Subject 6‡	1.00	1.04	0.56	0.59

^†Selective serotonin reuptake inhibitors.

^‡Central nervous system stimulants.

MEM: monocular estimation method.

Discussion

Mean differences between retinoscopic techniques and autorefraction spherical equivalent and between retinoscopic techniques and autorefraction power in the 180 meridian were all less than 0.14 D in both the children and adults in this study sample. The 95 per cent LOA were narrowest for comparisons to autorefraction power in the 180, ranging from ±0.80 to ±1.22 D. This level of agreement compares well with previous studies reporting agreement between retinoscopy and autorefraction for the measurement of accommodative accuracy in young adults,^9,10 and also with one previous study including some school- aged children.⁸ A comparison of these findings to past literature is shown in Table 3.

This study did not find the poor agreement between autorefraction and retinoscopic tests in children previously reported by the PEDIG study.¹¹ However, it should be noted that, unlike the present study, the PEDIG study specifically recruited children with high accommodative lags on autorefraction. As seen in Figure 1, the present study had very few children with accommodative lags greater than 1.00 D. However, the present study had several adult subjects with accommodative lags greater than 1.00 D whose agreement between techniques fell within ±0.50 D (Figure 1). In addition, a linear relationship between the magnitude of accommodative lag and magnitude of difference between tests was not observed in the present study. Differences in testing conditions across techniques may have led to the poorer overall agreement in the PEDIG study, and thus matching conditions (binocular viewing only, consistent room illumination, same target across methods, and use of subjects’ habitual corrections) may account for the better agreement that was observed with autorefraction in this study.

The present study also sought to determine whether calculating power in the 180 meridian from the autorefraction measures would improve agreement between autorefraction and retinoscopy techniques. There was a sizeable improvement for children in that a greater percentage of subjects had agreement within ±0.50 D when comparing autorefraction power in the 180 to the retinoscopic techniques (27 per cent improvement for MEM and 12 per cent for Nott). An improvement of 15 per cent was also seen in adults for Nott retinoscopy, but no improvement occurred for MEM in adults. However, the adults in the present study had significantly less uncorrected astigmatism compared to the children, which is a likely source of the difference in effect for children and adults.

While it could be debated whether agreement of ±0.50 D is clinically acceptable, this level of agreement was chosen for the descriptive analysis in this study based on previously published co-efficients of repeatability for retinoscopic versus autorefraction tests of accommodative accuracy.⁸ The study performed by PEDIG also reported the percentage of subjects whose measures had agreement of ±0.50 D (Nott versus autorefraction = 46 per cent, MEM versus autorefraction = 51 per cent),¹¹ which are much smaller percentages than observed in the present study for children.

One would predict that the calculation of power in the 180 meridian for comparison of autorefraction to MEM or Nott retinoscopy would be even more critical for agreement of measurements in subjects with higher amounts of uncorrected astigmatism. In this study, subjects with > 1.25 DC on aided autorefraction were excluded, which limited the range of uncorrected astigmatism available for analysis. Given that accommodative accuracy is typically assessed clinically with appropriate refractive error correction in place, the inclusion of subjects with larger amounts of uncorrected astigmatism was not thought to be generalisable to standard clinical procedure.

However, even with small amounts of uncorrected astigmatism, the present data suggest that the decision to utilise autorefraction spherical equivalent measures versus power in the 180 could impact agreement between techniques. This may also underscore the importance of fully correcting refractive error prior to assessing accommodative accuracy, or measuring the two principal meridians in uncorrected patients with astigmatism for whom the clinician is weighing whether or not to prescribe a correction.

While the mean difference in lag was 0.14 D or less across all comparisons between techniques, the individual differences between techniques reached up to 1.50 D for some subjects. One potential, albeit small, source of differences between techniques could be the inherently different measurement scale for each test. MEM was restricted to 0.25 D steps, Nott retinoscopy was measured on a continuous scale, and autorefraction was set to 0.12 D steps, although the use of an average of five measures made the autorefraction scale finer than 0.12 D.

The use of five repeated measures for autorefraction versus a single measure with both retinoscopic techniques represents a limitation of the present study; however, the goal of this work was to remain consistent with the common clinical technique. In addition, the autorefractor used in this study utilises a 2.4 mm measurement ring centred in the pupil to determine refractive error, whereas the retinoscopy techniques sample across the entire available pupil diameter. Thus findings may be influenced by differences in measurement principles. Despite these differences, the overall agreement found in the present study is consistent with past studies.

Conclusions

These findings demonstrate that agreement between retinoscopic and autorefraction techniques for the measurement of accommodative accuracy is similarly good in children and adults when matched testing conditions are used between techniques. In addition, these findings suggest that calculating lag for the meridian corresponding to that evaluated with retinoscopy may improve agreement between techniques even for subjects with moderately small amounts of uncorrected astigmatism.